Abstract:Large Language Models (LLMs) have recently displayed their extraordinary capabilities in language understanding. However, how to comprehensively assess the sentiment capabilities of LLMs continues to be a challenge. This paper investigates the ability of LLMs to detect and react to sentiment in text modal. As the integration of LLMs into diverse applications is on the rise, it becomes highly critical to comprehend their sensitivity to emotional tone, as it can influence the user experience and the efficacy of sentiment-driven tasks. We conduct a series of experiments to evaluate the performance of several prominent LLMs in identifying and responding appropriately to sentiments like positive, negative, and neutral emotions. The models' outputs are analyzed across various sentiment benchmarks, and their responses are compared with human evaluations. Our discoveries indicate that although LLMs show a basic sensitivity to sentiment, there are substantial variations in their accuracy and consistency, emphasizing the requirement for further enhancements in their training processes to better capture subtle emotional cues. Take an example in our findings, in some cases, the models might wrongly classify a strongly positive sentiment as neutral, or fail to recognize sarcasm or irony in the text. Such misclassifications highlight the complexity of sentiment analysis and the areas where the models need to be refined. Another aspect is that different LLMs might perform differently on the same set of data, depending on their architecture and training datasets. This variance calls for a more in-depth study of the factors that contribute to the performance differences and how they can be optimized.

Linguistically Regularized LSTM for Sentiment Classification

Utterance-Based Audio Sentiment Analysis Learned by a Parallel Combination of CNN and LSTM.

Sentiment Lexicon Enhanced Attention-Based LSTM for Sentiment Classification

Modality-invariant Temporal Representation Learning for Multimodal Sentiment Classification

A Mixture Language Model for the Classification of Chinese Online Reviews.

Audio Sentiment Analysis by Heterogeneous Signal Features Learned from Utterance-Based Parallel Neural Network.

SentiLR: Linguistic Knowledge Enhanced Language Representation for Sentiment Analysis

Encoding Syntactic Knowledge in Neural Networks for Sentiment Classification.

Sentiment Commonsense Induced Sequential Neural Networks for Sentiment Classification.

LAAN: A Linguistic-Aware Attention Network for Sentiment Analysis

Sentiment classification using bidirectional LSTM-SNP model and attention mechanism

An Augmented Neural Network for Sentiment Analysis Using Grammar

Learning Latent Opinions for Aspect-level Sentiment Classification

Sentence-Level Sentiment Classification A Comparative Study Between Deep Learning Models

Improving sentiment analysis via sentence type classification using BiLSTM-CRF and CNN

Sentiment Lexicon Enhanced Neural Sentiment Classification

Do Large Language Models Possess Sensitive to Sentiment?

A Deep Neural Architecture for Sentence-level Sentiment Classification in Twitter Social Networking

Student sentiment classification model based on GRU neural network and TF-IDF algorithm

Sentiment Analysis Based on LSTM Architecture with Emoticon Attention.

Attention-based LSTM for Aspect-level Sentiment Classification.